Service 03 / 04 · Brand models

Brand-trained AI models for images, product shots and text.

We train image, product and writing models on your brand, for owners and marketing heads. We use only files you have the rights to. A named person approves what goes live.

Start a brand AI project What it covers
Typical length
6–10 weeks to a trained model
Model types
Image · product · writing
Ownership
Yours once paid for

Illustration: a Flux base model with a brand add-on was trained on 1,240 approved files. It gets the request "autumn window, ceramic range, three versions" and makes three. A brand check scores each on colour, fonts and logo spacing. A (0.93) and B (0.90) pass and are approved. C fails colour and logo spacing at 0.78 and is sent back to be remade.

What we make

Six parts of a brand model, each available alone.

Training is a small part. Most work goes into the examples, rights, testing and upkeep.

  1. 01 / 06

    Training examples you may use

    We pick and label your files and check you may use each one. We keep some aside for testing.

    Labelled · consented

  2. 02 / 06

    Brand-trained image & product models

    Flux, Firefly or open models trained on your brand. New images match your style and show products as they are.

    Base + brand add-on

  3. 03 / 06

    Brand writing models

    Tested prompts or a trained model that keeps your voice, avoids your banned words and works in your writing tools.

    In your voice

  4. 04 / 06

    Brand match testing

    Fixed test examples for colours, logo use, fonts, layout and tone. They run on every model version before release.

    Scored every version

  5. 05 / 06

    Automatic checks & approval

    Rules on what models may make and what a person approves. We test them for misuse and hidden-instruction tricks first.

    Named approver

  6. 06 / 06

    Hosting, versions & care

    We host the models, track each version, set cost and speed limits and retrain when your brand changes.

    Version list · retrain

How it works

Four stages to a live model.

Your guidelines become training examples and tests. Your brand lead approves each model that passes. Timings are typical, set per project.

  1. Wk 01–03

    Define

    We label and rights-check your files, plus examples to avoid. Then we write the tests with your brand team.

    • Training examples
    • Rights list
    • Brand-match tests
  2. Wk 03–06

    Build

    We train a few models and test them on unseen examples. Your brand lead approves the best one.

    • Model options
    • Score report
    • Model notes
  3. Wk 05–08

    Test

    We add automatic checks, approvals and a log. We test against misuse and hidden-instruction tricks before launch.

    • Automatic checks
    • Misuse test report
    • Approval steps
  4. Wk 08–10

    Run

    We connect the model to your design and other tools. You get a version list, cost caps and a retrain plan.

    • Hosted model
    • Version list
    • Care plan

How we use AI

A model goes live only if it beats the untrained model on your brand.

We compare the trained model with the one it started from. Each check has a pass mark, agreed with your brand team first.

Model notes · Brand image model v3Release check
Model notes · Brand image model v3: illustrative sample
CheckUntrainedTrained v3Pass markResult
Colour difference (avg ΔE 2000) 5.81.7≤ 2.0Pass
Follows brand rules 46%88%≥ 85%Pass
Product shape accuracy 0.710.93≥ 0.90Pass
Unsafe or off-limits images 1.9%0.3%≤ 0.5%Pass
Readable text in images 38%61%≥ 80%Send to designer
Training images of known source
100%Licensed or owned
Cost per approved file
Index 45Untrained model = 100
Checks passed
4 / 5A designer does the fifth

Illustrative Screens marked Illustrative are examples, not client results.

What changes

More drafts that fit your brand, with the cost per file measured.

We measure every version against your colours and rules. Figures are typical; we set yours per project.

01

Your look from the start

The model is trained on your brand, so drafts start with your colours, fonts and layout.

First-time brand match

  • Before40%
  • Target85%+

Measured by Brand check on every version made

02

Colours that match yours

We measure how close each image is to your brand colours.

Colour difference (ΔE)

  • Before5.8
  • Target≤ 2.0

Measured by Mean ΔE 2000 against brand colours

03

Cost you can track

We add up model cost and review time for each file you approve.

Cost per approved file

  • BeforeIndex 100
  • TargetIndex 45

Measured by Model cost + review time, indexed

What you get

Your trained models, files included.

You get the model files, training data, rights records, tests and hosting settings. We never train other models on your data.

Colour swatch cards fanned into a full circle on a grey surface
Your brand colourswhat every image is scored against

Handover list · Brand Models7 items

Brand AI Tools: what you get and its format
#What you getFormat
01Rights-checked training & test examplesData · rights list
02Trained model files or add-onsFiles · your accounts
03Model notes & training recordsDoc
04Brand-match tests & reportTests · report
05Automatic check & misuse test resultsTests · report
06Model link & version listLink · version list
07Output log & approval stepsDashboard · steps

Two tools, one name

AI Design makes the work; Brand Design checks it.

Brand AI Tools sits in both areas. Here, it means models that make work in your brand style. In Brand Design, it means tools that check any file, however it was made.

  1. Brand Design Brand guidelines

    Logo, voice, colours and rules.

  2. AI Design · Brand AI Tools · this page Brand-trained models

    Trained on your approved work to make new versions.

  3. Brand Design · Brand AI Tools Brand checks

    Every file scored against the guidelines, from any source.

  4. Your team Approval

    A named person approves, and the record is kept.

The two meet at testing: we train our models until they pass Brand Design’s checks.

Explore Brand Design’s Brand AI Tools

Technologies we work with

Built with standard tools, hosted where you choose.

We train, test and host brand models with these tools. Model files and settings can move to any provider.

  • PyTorch
  • Hugging Face
  • Replicate
  • Modal
  • Python
  • MLflow
  • ONNX
  • NVIDIA
  • vLLM
  • Anthropic
  • Google Gemini
  • Figma
  • GitHub

Also in use

  • OpenAI

Standards we build to

AI and privacy standards applied to training and release.

They shape our training records, list of risks and release check. We build to them; we are not certified to them.

  • ISO/IEC 42001:2023Artificial intelligence management systems

    Rules for managing AI in a company: policy, impact checks and controls on data and models.

  • NIST AI RMF 1.0AI Risk Management Framework

    Managing AI risk in 4 parts: Govern, Map, Measure and Manage. A second guide covers generative AI (NIST AI 600-1).

  • EU AI ActArtificial Intelligence Act (EU) 2024/1689

    Europe's AI law sets rules by risk: banned uses, duties for high-risk AI and general-purpose models, and AI labels.

  • OWASP Top 10 for LLM ApplicationsSecurity risks in generative AI applications

    Risks in AI apps, such as hidden instructions that trick the AI, data leaks and AI with too much power.

  • GDPRGeneral Data Protection Regulation (EU) 2016/679

    Europe's data law: a legal reason to use data, people's rights, privacy by design, breach reports and risk checks.

  • DPDP Act 2023Digital Personal Data Protection Act, 2023

    India's data law and its Rules: notice and consent, people's rights, breach reports and extra duties for significant firms.

Offers & plans

Brand model services, from training data to ongoing care.

Buy one model, like an image or writing model, or the full build. Model files, data and logs go into your accounts. A named person approves what goes live.

Categories
04
Offers
14
Plans
04
Not sure what you need? Tell us the problem

How to buy

  1. 01Pick offers. Ask about one, or add several to one enquiry.
  2. 02Choose a plan. A small job, a fixed project or ongoing help.
  3. 03Send your enquiry. We reply within one working day.

Browse by category

Timelines are typical. We write the scope before we quote.

01Data & rights

3 offers
Typical timeline: 2–5 weeks

Training data you may use

We choose and label your files, then check the licence and consent for each one.

What’s included

  • Gathering, choosing and labelling files
  • Licence and consent check per item
  • Examples to avoid and tests
  • Version history and notes
  • Data
  • Rights-checked
  • Versioned

Best forAny brand about to train a model on its own files.

Typical timeline: 3–5 weeks

Brand knowledge base for AI

Your guidelines, rules and approved examples, set out so AI tools can read and quote them. We update them when your brand changes.

What’s included

  • Guidelines turned into clear rules
  • Searchable approved examples
  • Links to your tools
  • Update process for brand changes
  • Search
  • Sourced
  • Current
  • LlamaIndex
  • pgvector

Best forBrands whose rules sit in PDFs that no tool can read.

Typical timeline: 2–4 weeks

Model & tool selection

We pick the starting model and training method for each task by testing on your files, not public rankings.

What’s included

  • Shortlist of models for each task
  • Side-by-side test on your files
  • Licence, data location and cost review
  • Recommendation with reasons
  • Independent
  • Tested
  • Cost
  • OpenAI

Best forTeams comparing models and providers before they commit.

02Training

4 offers
Typical timeline: 4–8 weeks

Brand image model

An image model trained on your brand with Flux, Firefly or an open model. New images match your brand from the start.

What’s included

  • Light add-on or full training
  • Methods compared on unseen examples
  • Brand match score report
  • Files and prompts in your accounts
  • Base + brand add-on
  • Scored
  • Yours

Best forBrands making more images than anyone can check by hand.

Typical timeline: 4–8 weeks

Brand writing model

Tested prompts or a trained model that writes in your voice, uses your words and avoids banned ones.

What’s included

  • Voice and word list as rules
  • Tested prompts or a trained model
  • Scoring against the voice rules
  • Built into your writing tools
  • In your voice
  • Writing
  • In your tools
  • OpenAI

Best forTeams writing a lot across markets, channels and agencies.

Typical timeline: 5–9 weeks

Product photo model

A model trained on your products, so new scenes show them as they are: shape, materials, labels and finish.

What’s included

  • Product photo shoot or file review
  • Likeness training and checks
  • Scoring for product accuracy and style
  • Prompts for scenes and shots
  • Likeness
  • Product shots
  • Accuracy

Best forRetail and consumer brands with large catalogues and short seasons.

Typical timeline: 4–8 weeks

Style add-on library

Small, swappable add-ons for sub-brands, campaigns or markets. One model makes several looks without full retraining.

What’s included

  • Add-on per sub-brand, campaign or market
  • Rules for combining add-ons
  • Scoring per add-on
  • Names, versions and notes
  • Add-ons
  • Sub-brands
  • Swappable

Best forGroups running several brands from one team.

03Testing & approval

3 offers
Typical timeline: 3–5 weeks

Brand match testing

Agreed scores for colours, logo use, fonts, layout and tone, so release decisions no longer rest on opinion.

What’s included

  • Rules set with your brand team
  • Test examples kept out of training
  • Automatic scores matched to human review
  • Score report for each version
  • Testing
  • Re-tests
  • Per version

Best forBrands with lots of AI content and no clear test for it.

Typical timeline: 2–5 weeks

Automatic checks & misuse testing

Limits on what the tools may make. We test them against hidden-instruction tricks, rival logos, named people and banned content.

What’s included

  • Input and output checks
  • Tests from OWASP’s AI risk list
  • Blocked logos, people and claims
  • Findings, fixes and a retest
  • Automatic checks
  • Misuse tests
  • OWASP AI

Best forAny brand giving an AI tool to a wider team.

Typical timeline: 3–6 weeks

Approval steps & output log

A person approves AI work before it is used. We record the prompt, model version, inputs and approver for each output.

What’s included

  • Approval queue with named owners
  • Version and settings logged per output
  • Searchable output log
  • Export for audits or rights checks
  • Named approver
  • Full record
  • Slack

Best forRegulated brands, or teams where lawyers approve the work.

04Hosting & care

4 offers
Typical timeline: 4–8 weeks

Model hosting & versions

We host your trained models behind an API, a link other tools can use. We track versions, cap costs and can undo a release.

What’s included

  • Secure link for other tools
  • Version list and numbered releases
  • Cost, usage and speed limits
  • Undo plan and incident guide
  • API
  • Versioned
  • Cost limits

Best forTeams with a trained model only one person can run.

Typical timeline: 4–8 weeks

Built into your design tools

We add the models to your design tool as a plugin and to your website editor. Other systems use a secure link.

What’s included

  • Design tool plugin or app
  • Links to content and campaign systems
  • Secure log-in and team permissions
  • How-to guides for each team
  • Figma
  • Websites
  • API

Best forTeams who will not use a tool in a separate tab.

Typical timeline: Monthly, ongoing

Model care & updates

We retrain when your brand changes, score each release and check cost and brand match monthly.

What’s included

  • Planned retraining and re-scoring
  • Training data kept up to date
  • Watching cost, speed and brand match
  • Monthly report and to-dos
  • Retainer
  • Retrain
  • Monitored

Best forBrands that now rely on a trained model every day.

Typical timeline: 2–4 weeks

Handover to your team

Model files, data, scripts, scores and notes handed over, so your team can run and retrain the models itself.

What’s included

  • Your model files, data and scripts
  • Model notes and training records
  • Guide to retraining and release
  • Working sessions with your team
  • Handover
  • Documented
  • Ownership

Best forOrganisations taking model work in-house after a first build.

Your enquiry

Tick “Add to enquiry” on any offer, choose a plan, then continue. Or enquire about one offer.

How we work with you

Ways to start, from a question to a quote.

Ask a quick question, send your project details or request a quote. The person who will lead the work reads each one in full. Services in your enquiry go with it.

Or book a 30-minute call

What are you sending?

  1. 01

    About 2 minutes3 short answers

    For a first chat, a press request or anything not ready to plan yet.

    You get A reply from a named lead

  2. 02Recommended

    About 8 minutes5 short steps

    Goals, who it is for, budget and timing, so our first reply can outline the work.

    You get Options and prices after one call

  3. 03

    About 15 minutesYour documents attached

    Your documents, deadlines and the purchasing and security rules to meet.

    You get Confirmation and a named lead

How it is priced

Each plan shows how it is priced. Work starts once we agree a written scope and price.

  • Typical length
    1–3 weeks
    Pricing
    Fixed fee

  • Typical length
    4–12 weeks
    Pricing
    Fixed price

  • Typical length
    3–9 months
    Pricing
    Fixed price per stage

  • Typical length
    Ongoing · 6-month minimum
    Pricing
    Monthly fee
Compare what each plan includes
What each plan includes and who it suits
PlanWhat you getBest for
SprintA short, fixed job that answers one clear question.
  • Scope and result agreed before we start
  • A senior lead and the right specialists
  • A weekly review with you
  • A clear answer or a test version
 An audit, a brand check or a test version of an app
ProjectOne clear job for one fixed price.
  • A written scope: what you get and how we agree it is done
  • A named project lead and a fixed team
  • A shared plan with checkpoint dates
  • Source files, yours once paid for
 A brand identity, a website, a simple app or a set of tools
MilestoneA bigger build in stages you approve and pay for one at a time.
  • Stages, each with its own scope and approval
  • A go or stop review at each stage
  • Re-plan between stages as you learn
  • Pay for each stage you approve
 Work too big for one contract, such as a large app
RetainerA monthly team that runs and improves what we built.
  • Team time set aside for you every month
  • Agreed response times for requests and fixes
  • A monthly review and a to-do list
  • Planned improvements, not only fixes
 Monthly social media and ads, or care for live sites and apps

Questions

What buyers ask about brand models.

Answers on data, ownership, brand changes and misuse.

Enquire
How does this differ from Brand Design’s Brand AI Tools?

This one focuses on the models themselves. We prepare the data, then train, test, host and retrain them. Brand Design’s version adds the brand check and templates. You can start with either.

How much material do we need to train a model?

Less than you expect for style, more for likeness. A style add-on can learn from a few dozen labelled images. Products and people need varied, controlled photos. We test a small set first and say if it is not enough.

Who owns the trained model?

You do, including model files, add-ons, data, prompts and logs. What we make for you is yours once it is paid for. Tools we already had stay ours, and you get a free, permanent licence to use them.

What happens after launch?

Each model comes with a retraining plan and optional care from us. When your brand changes, we retrain on new material and test against old and new rules. The old version stays available until the new one passes.

Could it copy a real person or a rival’s brand?

No. Automatic checks block other brands’ logos and named people without recorded consent. All output is logged and every release is tested for misuse.

Also in AI Design

Other AI Design services that build on these models.

Each uses the models, test examples and approval rules built here.

All of AI Design

02 / 04 · Images and films

AI Content Studio

Visuals, films and voice-overs made with AI, with usage rights recorded.

Explore AI Content Studio

01 / 04 · AI assistants, agents

AI Application Design

AI assistants designed to show sources and pass hard questions to a person.

Explore AI Application Design

Tell us what you need built.

You will talk to the lead who would run your work.

Book a call

Three ways to start

We always agree a written scope and price before work begins.

Choose one of the three ways above
See the work Explore our services Industries we serve Meet the team Email us Read our approach